APU faster than GPU?

[Gikaseixas]

tential from your own links, i just copied the 1st part of performance test. AMD older tech (28nm vs 20nm) and lower clock speeds (1.0ghz vs 2.0ghz) does really well against Intel here. I'm just lazy to copy the 2nd part but its pretty much an embarrassment for AMD. Last it is also pretty meaningless because if i wanna compress a video or something of the sort i'll do it on a proper desktop or laptop

AMD A6-1450
(4C/4T x 1.0 GHz, 28nm, 2MB L2, 8W TDP)

Power Consumption (Lower is better)
Youtube HTML5 Streaming - 8.82w
Youtube Flash Streaming - 8.73w
Netflix Streaming - 8.42w

Performance (Higher is better)
Futuremark PCMark 8 Home - 1293
Futuremark PCMark 8 creative - 1302
Futuremark PCMark 8 Work - 1825
Futuremark PCMark 7 - 2069
Futuremark 3DMark 11 - 757
Futuremark 3DMark 2013 Ice Storm - 16571
Futuremark 3DMark 2013 Cloud Gate - 1549

3D Rendering - CINEBENCH R15 (Higher is better)
Single threaded - 21
Multi threaded - 78
Open GL - 9.06

Intel Celeron J1900
(4C/4T x 2.00 GHz (2.41 GHz Turbo), 22nm, 2MB L2, 10W)

Power Consumption (Lower is better)
Youtube HTML5 Streaming - 8.36w
Youtube Flash Streaming - 8.02w
Netflix Streaming - 6.62w

Performance (Higher is better)
Futuremark PCMark 8 Home - 1330
Futuremark PCMark 8 creative - 1302
Futuremark PCMark 8 Work - 1390
Futuremark PCMark 7 - 2803
Futuremark 3DMark 11 - 463
Futuremark 3DMark 2013 Ice Storm - 15395
Futuremark 3DMark 2013 Cloud Gate - 1490

3D Rendering - CINEBENCH R15 (Higher is better)
Single threaded - 39
Multi threaded - 145
Open GL - 5.87

 

[AtenRa]

tential from your own links, i just copied the 1st part of performance test. AMD older tech (28nm vs 20nm) and lower clock speeds (1.0ghz vs 2.0ghz) does really well against Intel here. I'm just lazy to copy the 2nd part but its pretty much an embarrassment for AMD. Last it is also pretty meaningless because if i wanna compress a video or something of the sort i'll do it on a proper desktop or laptop

AMD A6-1450
(4C/4T x 1.0 GHz, 28nm, 2MB L2, 8W TDP)

Power Consumption (Lower is better)
Youtube HTML5 Streaming - 8.82w
Youtube Flash Streaming - 8.73w
Netflix Streaming - 8.42w

Performance (Higher is better)
Futuremark PCMark 8 Home - 1293
Futuremark PCMark 8 creative - 1302
Futuremark PCMark 8 Work - 1825
Futuremark PCMark 7 - 2069
Futuremark 3DMark 11 - 757
Futuremark 3DMark 2013 Ice Storm - 16571
Futuremark 3DMark 2013 Cloud Gate - 1549

3D Rendering - CINEBENCH R15 (Higher is better)
Single threaded - 21
Multi threaded - 78
Open GL - 9.06

Intel Celeron J1900
(4C/4T x 2.00 GHz (2.41 GHz Turbo), 22nm, 2MB L2, 10W)

Power Consumption (Lower is better)
Youtube HTML5 Streaming - 8.36w
Youtube Flash Streaming - 8.02w
Netflix Streaming - 6.62w

Performance (Higher is better)
Futuremark PCMark 8 Home - 1330
Futuremark PCMark 8 creative - 1302
Futuremark PCMark 8 Work - 1390
Futuremark PCMark 7 - 2803
Futuremark 3DMark 11 - 463
Futuremark 3DMark 2013 Ice Storm - 15395
Futuremark 3DMark 2013 Cloud Gate - 1490

3D Rendering - CINEBENCH R15 (Higher is better)
Single threaded - 39
Multi threaded - 145
Open GL - 5.87

A6-1450 is a TABLET SoC released in Q2 2013.

J1900 is a DESKTOP SoC released in Q1 2014.

Apples to Oranges

Also, anyone that needs 1080p/60 will not use either of them.

AT commend on J1900

The target market for this type of system will probably not be interested in anything other than vanilla 1080p24 content, and for those folks, the system can act as a good HTPC with EVR-CP as the renderer.

also,

and the most important

Another aspect that HTPC users need to keep in mind while considering Bay Trail systems is the HD audio bitstreaming capability. The hardware is indeed capable, and we do get Dolby Digital Plus bitstreaming with the Windows 8.1 Netflix app. However, XBMC and other commercial software Blu-ray players are unable to bitstream DTS-HD and TrueHD audio on Windows 8.1 due to a missing software component for the Bay Trail chipset. OpenELEC (and other such bootable operating systems) has no problems, though, because it doesn't rely on the Windows component (obviously) for HD audio support.

So, you can see 1080p/24fps on the Baytrail but you will not have DTS-HD or TrueHD audio.

But tential completely missed that and believes that A6-1450 users are going for 1080/60fps. :whiste:

 

[redzo]

If the memory bottleneck can be removed (with HBM, for instance), theoretically an APU can be faster than a similar dedicated GPU because the communication between CPU and GPU is better?

Sorry for the non technical words.

No offence taken: non-technical also.

How about market segmentation, market monopoly and customer needs,

All that matters is price and satisfying market demand. Technicalities are not important. We all fancy them because we are passionate about technology, but in the grand scheme of things they do not matter. All that matters is financial wisdom <=> profit.

So what if you can suddenly fit a r9 280x level GPU with no bottle necked memory to a CPU!
Nothing will change because it is all about customer needs. The problem is that a full dedicated 280x class gpu + equivalent intel cpu will be cheaper than your apu offering with that 280x on the cpu die. Once you fit a cpu + gpu on to the same die, you will take away a customer's right of choosing between a lot of options. You will never be capable of launching so many APU models in so many combinations in order to satisfy all your customer's needs. Your offering will be very limited. Don't forget that intel sells APU's also, but they've basically moved their northbridge graphics on to the cpu die not because it is fancy/cool, but because it makes financial wisdom.

They(intel) haven't reinvented the wheel. Intel's model works because their GPUs were crap and they are still crap, but free. They didn't care about not being able of satisfying demand because their gpu performance was crap and, let's not forget, they had monopoly on that free crap. They had market monopoly in terms of free(no cost) low performance GPUs(i'am talking about the northbridge ones) for everyone and they basically own(owned) that market.

Nobody complains about intel HD because it is free of charge. The burden of performance APU GPUs rests only on AMD's shoulders.

 

[tential]

A6-1450 is a TABLET SoC released in Q2 2013.

J1900 is a DESKTOP SoC released in Q1 2014.

Apples to Oranges

Also, anyone that needs 1080p/60 will not use either of them.

AT commend on J1900
also,

and the most important

So, you can see 1080p/24fps on the Baytrail but you will not have DTS-HD or TrueHD audio.

But tential completely missed that and believes that A6-1450 users are going for 1080/60fps. :whiste:

Intention or not, you CAN do it on Baytrail(without frame drops so why you would say if you need 1080p/60fps you can't use baytrail is beyond me), you absolutely can't on AMD's A6-1450 at the same pricepoint.

So a Desktop SoC consumes similar power to AMD's tablet SoC and performs better?
That's embarrassing.

So much for being a leader in APU technology.

 

[Gikaseixas]

So a Desktop SoC consumes similar power to AMD's tablet SoC and performs better?
That's embarrassing.

Performs better at what? Encoding? lol
did you noticed that you comparing a higher clocked SoC and one that was released 3 quarters later? Intel is dirty rich and this is the best they can do? A cash constrained AMD is keeping them honest at least in this context.

 

[tential]

Performs better at what? Encoding? lol
did you see that you comparing a higher clocked SoC and one that was released 3 quarters latter?

AMD executes yearly? According to Aeterna's "Fastest APU each year!!!!". So I compared the products available now. If AMD wants to release an APU that makes it to retail products in 2014 be my guest. It's not my fault they didn't.

So does AMD execute year after year or not?

 

[mrmt]

AtenRa showed you that AMD has been releasing APU's in a consistent manner for 4 straight years and as i said it before we're talking about an industry that needs to be taken with a grain of salt. So it's not an absolute, it's simply an expectation from me that AMD can deliver this one.

I don't read AtenRa's posts, so I can only wonder about how he might be trying to warp reality in order to say that AMD is releasing APUs in a consistent manner for 4 straight years. If you bothered to read the official roadmaps I linked here, this is what you were going to find:

- Llano was late

- Krishna and Wichita were canned

- Trinity was on time

- Richland wasn't supposed to exist

- Kaveri was supposed to be a 2013 product

- Carrizo was supposed to be a 2014 product

- There wasn't supposed to exist any Carrizo-L APU

- There was supposed to exist a Steamroller server APU

- There was supposed to exist an Excavator server APU

- Every single one of the big core APU was below disclosed performance forecast.

- The small core APUs were more or less on time after Krishna. The cat family is very much resilient, the work that started on Brazos was really top notch.


============

So no, AMD isn't consistently releasing APUs for the last 4 years. The only consistent things during the last 4 years at AMD are:

- They consecrate their roadmaps to the trash can every single year

- They lose market share

- They lose money

 

[Essence_of_War]

- There was supposed to exist a Steamroller server APU

Slated for 2014 release, currently MIA with no indication of when, if ever, it will exist.

 

[Abwx]

AMD executes yearly? According to Aeterna's "Fastest APU each year!!!!". So I compared the products available now. If AMD wants to release an APU that makes it to retail products in 2014 be my guest. It's not my fault they didn't.

So does AMD execute year after year or not?

You should have taken a bobcat as comparison, why using an A6 1450.?.

And instead of the J1900 you should have picked a 4770k.

Serious, and non biaised, people will look at decent reviewer to see wich product is better.

http://nl.hardware.info/reviews/5330/amd-am1-vs-intel-bay-trail-d-review-goedkope-desktopplatforms

http://www.computerbase.de/2014-04/amd-athlon-5350-kabini-sockel-fs1b-test/

So much for Anand s power figures and the J1900 alleged lower TDP, perfs i wont talk.

A6-1450 is a TABLET SoC released in Q2 2013.

J1900 is a DESKTOP SoC released in Q1 2014.

Cherry picky is still unbiaised in comparsion of this...comparison.

 

[AtenRa]

I don't read AtenRa's posts, so I can only wonder about how he might be trying to warp reality in order to say that AMD is releasing APUs in a consistent manner for 4 straight years. If you bothered to read the official roadmaps I linked here, this is what you were going to find:

- Llano was late

- Krishna and Wichita were canned

- Trinity was on time

- Richland wasn't supposed to exist

- Kaveri was supposed to be a 2013 product

- Carrizo was supposed to be a 2014 product

- There wasn't supposed to exist any Carrizo-L APU

- There was supposed to exist a Steamroller server APU

- There was supposed to exist an Excavator server APU

- Every single one of the big core APU was below disclosed performance forecast.

- The small core APUs were more or less on time after Krishna. The cat family is very much resilient, the work that started on Brazos was really top notch.


============

So no, AMD isn't consistently releasing APUs for the last 4 years. The only consistent things during the last 4 years at AMD are:

- They consecrate their roadmaps to the trash can every single year

- They lose market share

- They lose money

- Ivybridge was late

- haswell was late

-Broadwell IS late

-22nm was supposed to be 2011

-14nm was supposed to be 2013

- 2011 Mobile group lose money

- 2012 Mobile group lose money

- 2013 Mobile group lose money

- 2014 Mobile group lose money

- 2011. iGPU second best to AMD, next year will be 2x performance

- 2012. iGPU second best to AMD, next year will be 2x performance

- 2013. iGPU second best to AMD, next year will be 2x performance

- 2014. iGPU second best to AMD, next year will be 2x performance

- next year 2x performance

You see, goes both ways :whiste:

 

[AtenRa]

If AMD wants to release an APU that makes it to retail products in 2014 be my guest. It's not my fault they didn't.

So does AMD execute year after year or not?

There are many Beema products in 2014 to choose from, but you concentrated to 2013 Tablet SoC Kabini. :whiste:

 

[Essence_of_War]

....

You see, goes both ways :whiste:

These claims have never been in contention. No one has argued with you about the numbers of dollars Intel's mobile group has lost, no one has argued about whether or not 14nm was delayed.

The contention was whether your post above:

AMD
Llano Q3 2011, fastest APU in the market(desktop)
Trinity Q3 2012, fastest APU in the market(desktop)
Ritchland Q2 2013, fastest APU in the market(desktop)
Kaveri Q1 2014, fastest APU in the market(desktop)

FOUR CONSECUTIVE fastest and on time APU releases the last 4 years.

Now that is a track record of them being able to DELIVER and they will do the same with Carrizo.

painted an overly optimistic picture of "consistency" in product releases while glossing over all kinds of details to the contrary.

 

[tential]

There are many Beema products in 2014 to choose from, but you concentrated to 2013 Tablet SoC Kabini. :whiste:

Link me then since i didn't concentrate on any. Please stop playing the victim. I simply picked readily available configurations in that small nuc configuration.

Instead of playing the victim, post the correct retail available product.

 

[Phynaz]

- Ivybridge was late

- haswell was late

-Broadwell IS late

-22nm was supposed to be 2011

-14nm was supposed to be 2013

- 2011 Mobile group lose money

- 2012 Mobile group lose money

- 2013 Mobile group lose money

- 2014 Mobile group lose money

- 2011. iGPU second best to AMD, next year will be 2x performance

- 2012. iGPU second best to AMD, next year will be 2x performance

- 2013. iGPU second best to AMD, next year will be 2x performance

- 2014. iGPU second best to AMD, next year will be 2x performance

- next year 2x performance

You see, goes both ways :whiste:

Intel operating cash flow - $20.68B
AMD operating cash flow - (193.00M)

Execution. Something AMD has lacked for over a decade, maybe two decades.

Still waiting for your fastest APU justification btw.

 

[mrmt]

Still waiting for your fastest APU justification btw.

Perhaps he thinks we are as gullible as his customers.

 

[Gikaseixas]

If the memory bottleneck can be removed (with HBM, for instance), theoretically an APU can be faster than a similar dedicated GPU because the communication between CPU and GPU is better?

It is possible but i think it would end up around a similar level of performance. That if HBM can ever be implemented. If they succeed we might see improvements but if not...

P.S. Trying to get thread back on topic

 

[AtenRa]

Still waiting for your fastest APU justification btw.

Llano vs HD3000 Sandybridge review
http://www.anandtech.com/show/4476/amd-a83850-review/5

Complete domination in GPU and very competitive CPU Multithread performance in its segment(Core i3).

Trinity vs HD4000 Ivybridge review
http://www.anandtech.com/show/6332/amd-trinity-a10-5800k-a8-5600k-review-part-1/2

Complete domination in GPU/GPGPU and very competitive CPU Multithread performance in its segment(Core i3).

Richland vs HD4600 Haswell review
http://www.anandtech.com/show/7032/...-gpu-on-the-desktop-radeon-hd-8670d-hd-4600/2

https://translate.google.com/transl...art-1-desktop-applications&edit-text=&act=url

https://translate.google.com/transl...clocked-part-2-igpu-gaming&edit-text=&act=url

Complete domination in GPU/GPGPU and very competitive CPU Multithread performance in its segment(Core i3).

Kaveri vs HD4600 Haswell refresh review
http://www.anandtech.com/show/7677/amd-kaveri-review-a8-7600-a10-7850k/12

http://forums.anandtech.com/showpost.php?p=35956808&postcount=259
http://forums.anandtech.com/showpost.php?p=35956810&postcount=260

http://apusilicon.com/apu-wars-amd-a10-7700k-vs-intel-core-i3-4330/

Complete domination in GPU/GPGPU and very competitive CPU Multithread performance in its segment(Core i3).

Also all of them are completely faster in iGPU even against the High-End Core i5/i7 APUs in each era.

But you already knew all that didnt you ??

 

[AtenRa]

It is possible but i think it would end up around a similar level of performance. That if HBM can ever be implemented. If they succeed we might see improvements but if not...

P.S. Trying to get thread back on topic

The APU will be faster because of Huma and HSA.

 

[mrmt]

It is possible but i think it would end up around a similar level of performance. That if HBM can ever be implemented. If they succeed we might see improvements but if not...

P.S. Trying to get thread back on topic

I think the most important question regarding HBM is costs. Given that AMD products are mostly confined to the low cost market, will HBM be able to reach the low costs necessary to compete against dGPU solutions?

 

[Phynaz]

Llano vs HD3000 Sandybridge review
http://www.anandtech.com/show/4476/amd-a83850-review/5

Snip

From your first link:

http://www.anandtech.com/show/4476/amd-a83850-review/3

Not so fast is it? Did you bother to read the article conclusion before you posted?

If gaming isn't something you're going to be doing then you're better off with Sandy Bridge.

Wondering why you are using the 2x series to prove your point, where's the 4x? Why do you have to use a 3.5 year old review?

We knew you would cherry pick to make your point. In the big picture AMD doesn't have the fastest "APU". They continue to show an inability to execute on their road-map. This is born out by their market share and financial performance.

 

[AtenRa]

From your first link:

http://www.anandtech.com/show/4476/amd-a83850-review/3

Not so fast is it? Did you bother to read the article conclusion before you posted?

Wondering why you are using the 2x series to prove your point, where's the 4x? Why do you have to use a 3.5 year old review?

We knew you would cherry pick to make your point. In the big picture AMD doesn't have the fastest "APU". They continue to show an inability to execute on their road-map. This is born out by their market share and financial performance.

from the same link

also,

46fps vs 21fps = more than 2X faster

40fps vs 23 = almost 2X faster

40fps vs 26fps = 53% faster

34,6fps vs 10,9fps = 3X faster

65fps vs 28,5 = more than 2X faster

That is humiliation, Llano was the absolute APU of its time. Its evidence that having the faster product doesnt mean you will always sell more than your competitor

 

[ShintaiDK]

Its evidence that having the faster product doesnt mean you will sell more than your competitor

But it wasnt faster. It was faster in a few things, slower in most.

 

[AtenRa]

But it wasnt faster. It was faster in a few things, slower in most.

I will suggest to read the review again, even in CPU benchmarks it is equal to Core i3 Sandybridge in the AT review. And 2x or even 3x faster in gaming using the iGPU. That translates as being the Faster APU in its era.

 

[ShintaiDK]

I will suggest to read the review again, even in CPU benchmarks it is equal to Core i3 Sandybridge in the AT review. And 2x or even 3x faster in gaming using the iGPU. That translates as being the Faster APU in its era.

No its not.

http://www.anandtech.com/show/4476/amd-a83850-review/3

And nobody with half a mind buys an APU for its gaming performance. They buy CPU+GPU.

 

[AtenRa]

No its not.

http://www.anandtech.com/show/4476/amd-a83850-review/3


And nobody with half a mind buys an APU for its gaming performance. They buy CPU+GPU.

Whatever floats your boat.